[boost] Re: Compile time assertions

From: Dave Abrahams (abrahams_at_[hidden])
Date: 2000-02-10 21:58:23

on 2/10/00 7:52 PM, Beman Dawes at beman_at_[hidden] wrote:

Just my 2 cents on these, FWIW. I'm tired of discussions that flail around
trying to fulfill an unrealistic set of impossible goals.

> (R1) Give an error message *at the line* where the assertion
> fails.

Highly compiler-dependent! No single implementation is *ever* likely to
satisfy this requirement! Also, some people want readability...
CodeWarrior's error messages are often so dismal that it can be very hard to
accomplish anything that looks reasonable in this regard. On the up side,
either class or function templates (I forget which) do give an instantiation
backtrace when they fail to compile.

> (R4) No macros.

What is the reason for this requirement? Just that we don't like macros? I
agree with the sentiment, but let's not be arbitrarily draconian. We use
them for #include guards. We use good old assert(); it works quite well, and
is always implemented with a macro.

Is it more important to stand our ground on principle, or to produce a
useful library component?

Okay, maybe that's more like $2.
<flame off>